Artificial Intelligence in the Boardroom
In recent years, there has been an increasing number of reports in the business press about foreign corporations deploying AIs in leadership roles. In his dissertation, Karl Döding, senior research fellow at the Institute, explores whether the use of an AI on a corporate board would require changes to German corporate law, and if so, what those changes would have to be.
The current race to build more and more powerful AI seems to presage a time, in the not-too-distant future, when AI agents in a corporation will be capable not only of following instructions, but of making decisions in the name of the organization itself. An AI as a corporate manager would be authorized to issue instructions that the human employees of the corporation would be obliged to carry out. For the success of the enterprise, the quality of the technology would mean the difference between swimming and going belly-up. Currently, German law does not allow an AI to be deployed as a member of the management board. On the contrary: German corporate law mandates that the management of a German stock corporation (an AG) must be human. But if other jurisdictions move to permit AIs to be deployed in corporate leadership roles, international regulatory competition would create a chain-shift effect, forcing German policy makers to contemplate changes in domestic law.
Döding calls the figure of an agentic AI, deployed on the management board and therefore charged (or should we say, prompted) with leadership duties, a Leitungsagent or “AI director”. Döding’s forward-looking study explores the challenges such AI directors pose to corporate law, and to the corporate regulatory landscape in general, in which the technology may be new, but the contours are familiar, involving the principal-agent conflict between shareholders and directors, the internal tension between controlling and minority shareholders, and the conflict between shareholder and stakeholder interests, including the general public.
Döding’s research shows that this “solution” to the principal-agent conflict within the corporation is no solution at all. The emergence of the AI director merely shifts this classic conflict onto a more technology-driven plane. The same applies to the conflict between controlling and minority shareholders: the AI director functions like a ball that’s always in the controlling shareholder’s court, adding a whole new dimension to the conflict. As far as stakeholder interests are concerned, questions arise about who should bear the risk of malfunctions and about what goals the AI director should be prompted to pursue in the first place. Döding ends his presentation with thoughts on corporate law as a behavioral guiderail for the corporation and concludes that the current steering mechanisms are not equipped to guide AI decision-making. For lawmakers seeking to constrain these developments, what comes to the fore as a possible fulcrum is the development and training of AI systems, which are significant processes for all stakeholders.
Dr Karl Döding studied law the University of Münster and during that time worked at the university’s Centre for Insurance Studies. His dissertation was supported by a doctoral fellowship from the Konrad Adenauer Foundation. His dissertation won the Münster Law School’s dissertation prize and an award from the Esche Schümann Commichau Foundation.
Image: © Max Planck Institute for Comparative and International Private Law / Bastian Kurzynsky
